From 691a5da746226ddc630ec1e1d419aa54e3cec7cb Mon Sep 17 00:00:00 2001
From: Joseph Magly <1159087+jmagly@users.noreply.github.com>
Date: Thu, 27 Aug 2026 20:17:09 -0400
Subject: [PATCH] fix(qwen): recognize saved text derivatives

---
 obliteratus/models/loader.py      |  8 ++++++--
 tests/test_app_model_lifecycle.py | 10 ++++++----
 tests/test_loader_boundaries.py   | 15 +++++++++++++++
 3 files changed, 27 insertions(+), 6 deletions(-)

diff --git a/obliteratus/models/loader.py b/obliteratus/models/loader.py
index 4103813..c6eb0c4 100644
--- a/obliteratus/models/loader.py
+++ b/obliteratus/models/loader.py
@@ -666,7 +666,7 @@ def qwen_hybrid_runtime_overrides(
     contract. Generic Accelerate sharding is not valid for the recurrent
     DeltaNet execution path, even when aggregate VRAM is sufficient.
     """
-    if getattr(config, "model_type", "") != "qwen3_5":
+    if getattr(config, "model_type", "") not in {"qwen3_5", "qwen3_5_text"}:
         return {}
 
     import transformers
@@ -876,7 +876,11 @@ def load_model(
     if task == "classification":
         config.num_labels = num_labels
         load_kwargs["config"] = config
-    is_qwen_hybrid = task == "causal_lm" and getattr(config, "model_type", "") == "qwen3_5"
+    is_qwen_hybrid = task == "causal_lm" and getattr(
+        config,
+        "model_type",
+        "",
+    ) in {"qwen3_5", "qwen3_5_text"}
     qwen_device_index = None
     if is_qwen_hybrid:
         qwen_overrides = qwen_hybrid_runtime_overrides(
diff --git a/tests/test_app_model_lifecycle.py b/tests/test_app_model_lifecycle.py
index f36d6f7..408a103 100644
--- a/tests/test_app_model_lifecycle.py
+++ b/tests/test_app_model_lifecycle.py
@@ -113,15 +113,17 @@ app._load_model_to_device(checkpoint, local_files_only=True)
 assert calls[0][0] == checkpoint
 assert calls[0][1]["local_files_only"] is True
 
-loader.qwen_hybrid_runtime_overrides = Mock(
-    return_value={"attn_implementation": "sdpa", "device_map": {"": 2}},
+app.AutoConfig.from_pretrained.return_value = SimpleNamespace(
+    model_type="qwen3_5_text",
 )
+loader._require_qwen_hybrid_kernels = Mock()
+loader._estimate_model_memory_gb = Mock(return_value=54.0)
+loader._qwen_single_device = Mock(return_value=2)
 app._load_model_to_device(checkpoint, local_files_only=True)
 assert calls[-1][1]["attn_implementation"] == "sdpa"
 assert calls[-1][1]["device_map"] == {"": 2}
 assert calls[-1][1]["device_map"] != "auto"
-loader.qwen_hybrid_runtime_overrides.reset_mock(return_value=True)
-loader.qwen_hybrid_runtime_overrides.return_value = {}
+app.AutoConfig.from_pretrained.return_value = SimpleNamespace(model_type="gpt2")
 
 quantization = object()
 loaded_model, loaded_tokenizer = app._reload_local_checkpoint(
diff --git a/tests/test_loader_boundaries.py b/tests/test_loader_boundaries.py
index 831f713..d63140d 100644
--- a/tests/test_loader_boundaries.py
+++ b/tests/test_loader_boundaries.py
@@ -573,6 +573,21 @@ def test_qwen_single_device_uses_per_device_free_memory(monkeypatch):
     assert loader._qwen_single_device(216.0, "4bit") == 1
 
 
+def test_qwen_saved_text_derivative_reuses_single_device_runtime(monkeypatch):
+    config = _config(model_type="qwen3_5_text")
+    monkeypatch.setattr(loader, "_require_qwen_hybrid_kernels", Mock())
+    monkeypatch.setattr(loader, "_estimate_model_memory_gb", Mock(return_value=54.0))
+    monkeypatch.setattr(loader, "_qwen_single_device", Mock(return_value=2))
+
+    assert loader.qwen_hybrid_runtime_overrides(
+        config,
+        torch.bfloat16,
+    ) == {
+        "attn_implementation": "sdpa",
+        "device_map": {"": 2},
+    }
+
+
 def test_model_handle_metadata_snapshot_restore_summary_and_cleanup(tmp_path):
     model = _model()
     nested = SimpleNamespace(